bobye

It depends on Breeze

UFLDL Tutorial Implementation

1

NNet Models - Template + Module based Framework - Feedforward and back-propagation (BP) - Sparse Autoencoder - Stacked Autoencoder Utilities - Objective templates: L1/L2 norm, softmax, ... - Training methods:...

parallel gradient of unfolding RAE

1

error in BP: mem(key).inputBuffer() has inconsistent length

Good features to track

3

最近在做tracking相关的工作，这篇文章拿出来先看。

enable the modularity of core API

9

1. Need to remove dependency on explicitly computing the pruning weights. (Current is hard-coding) 2. Need to properly handle input.op.name in compressed mode. (Current is hard-coding)